Interesting, that should work just fine then. Locally I am having the same results as you, the message queue grows, but as in your example, it doesn’t go over 20-30MB and then it enters into discard. Notice you cannot look at the mode in your log results because they are quite delayed. Instead, you can check the Logger.Config table in Observer.
In any case, I have been thinking about this and I may have come up with a better approach for doing back-pressure that will be a bit more resilient to the handler blocking, I will follow up here once I have a POC.
